We searched for candidate Wnts expressed near prehaematopoietic mesoderm and identified a previously uncharacterized zebrafish wnt16 orthologue (Fig. S1) expressed in the dorsoanterior portion of more rostral somites from 10 hours post-fertilization (hpf; tailbud-stage) to 24 hpf (Fig. S2), the time frame when prehaematopoietic mesoderm transitions to recognizable HSC precursors in the dorsal aorta. Wnt16 is conserved across vertebrate phyla (Fig. S1), and the human form was originally identified as a gene inappropriately expressed in pre-B-acute lymphoblastic leukaemia (ALL) cells containing the E2A-PBX1 t(1;19) translocation product⁶. In zebrafish, two wnt16 splice variants are produced, only one of which is active (Fig. S1, a, b; Fig. S3; Table S1).

Knock down of Wnt16 by injection of either of two “splice-blocking” morpholinos caused a striking haematopoietic phenotype. Morpholinos alone or in combination reduced functional wnt16 message (Figs. S1, a; S4), yielding highly similar phenotypes in comparison to a 5-basepair mismatch control morpholino, which had no effect. (Figs. S4, S5). HSC precursors, a population of runx1⁺ cells in the ventral floor of the dorsal aorta^{1, 7, 8}, as in mouse⁹, are absent in embryos injected with W16MO at 24 hpf by whole mount in situ hybridization (WISH; red arrowhead; Fig. 1 a, b; Table S2). By 33 hpf, HSCs can be observed as a population of cmyb⁺ cells between the dorsal aorta and posterior cardinal vein¹, and this population is also absent (red arrowhead, Fig. 1 c, d; Table S2). HSCs observable in living transgenic animals carrying GFP under the control of the cd41 (itga2b) promoter^10-12 are absent or reduced in wnt16 morphants (red arrowheads, Fig. 1 e, f; Movies S1 and S2; Table S2), as are unique double-positive cells in kdrl:RFP and cmyb:GFP double transgenics⁷ (Fig. 1 g, h, yellow cells; Table S2), whereas unrelated GFP-labelled multi-ciliate cells in the pronephros are unaffected (yellow arrowheads, Fig. 1 e, f; green cells, Fig. 1 g, h). Generation of T-lymphocytes requires an HSC precursor^{10, 11, 13}, providing a useful readout for whether HSCs have been specified or not. In cd41:GFP transgenic animals, GFP⁺ HSCs retain residual fluorescence as they differentiate into the first thymic lymphocytes around 3 days post fertilization (dpf)¹⁰, and these thymic immigrants are greatly reduced in wnt16 morphants (Movies S3 and S4). Accordingly, wnt16 morphants lack rag1⁺ T lymphocytes at 4.5 dpf (blue arrows, Fig. 1, i, j; Table S2), despite intact thymic epithelium (Fig. S6). Together, our results indicate that Wnt16 is required for specification of HSCs during embryonic development.

To determine the specificity of W16MO-induced defects, we examined the integrity of non-hematopoietic tissues by morphology and expression of marker genes (Fig. 2; Table S3). Wnt16 morphants are grossly normal (Fig. 2 a, b), with intact and functional vasculature, as evidenced by beating hearts and circulating primitive, non-HSC-derived erythroid cells, although there appear to be subtle patterning defects in the number and position of intersegmental vessels. By WISH, W16MO-injected embryos have primitive blood (gata1; Fig. 2 c, d; Table S3), somites (myod; Fig. 2 e, f; Table S3), vasculature (tll1, cdh5, and flk1; Fig. 2 g-l; Table S3), dorsal aorta (efnb2a; Fig. 2 m, n; notch1b, notch3, dll4 Fig. S7, a-d, o, p; Tables S3, S5), hypochord (col2a1a; Fig. 2 o, p; Table S3), notochord and floorplate (shha; Fig. 2 q, r; Table S3), and pronephros (cdh17; Fig. 2 s, t; Table S3). Thus, defects in HSC specification in wnt16 morphant animals are highly specific and not due to wholesale failure in specification of nearby tissues.

Canonical, β-catenin/Tcf-dependent Wnt signalling has been reported to be involved in HSC specification in mouse² and very early maintenance in zebrafish⁵. We therefore wanted to determine if Wnt16 is canonical. Overexpression of Wnt16 caused phenotypic defects demonstrating active protein (Fig. S3), but in comparison to Wnt3¹⁴, did not cause ectopic expression of canonical targets (Fig. S8). Conversely, Wnt16 knock down caused no decrease in canonical reporter activity (Fig. S9). Thus, Wnt16 does not act via β-catenin/Tcf and must act via a non-canonical pathway as has been suggested for human WNT16b^{15, 16}. Loss of other non-canonical Wnt ligands causes distinct phenotypes. Loss of Wnt5b is much more severe causing nearly complete absence of primary trunk vasculature¹⁷, whereas loss of Wnt11 had no effects on HSC specification (Fig. S10, Table S4). We therefore conclude that the HSC defects caused by loss of wnt16 are not a general consequence of loss of non-canonical Wnt signalling.

Notch signalling is required across phyla for developmental specification of HSCs¹. Global inhibition of Notch signalling by mutation or targeted deletion of Mindbomb (Mib), which is required for Notch ligand activity, or the essential Notch DNA-binding co-factor Rbpjκ results in total loss of HSCs, as well as severe vascular defects¹. If Wnt16 were to regulate Notch pathway genes, this regulation might explain the loss of HSCs in wnt16 morphant animals. The specific ligands and receptors required in zebrafish for specification of HSCs by the Notch pathway have not been determined. We therefore examined comparative expression of Notch ligands and receptors that might participate in HSC specification in uninjected or W16MO-injected animals. Most Notch receptors and ligands were either unaffected or very weakly affected (Fig. S7; Fig. 4 c,d; Table S5). However, expression of two Notch ligands, dlc and dld, was markedly decreased in somites at 17 hpf (16-ss; Fig. 3 a-d, Table S5). In accord with decreased somitic ligand expression, Notch reporter activity was decreased in somites at 17.5 hpf (Fig. S11). Interestingly, diminution of dlc and dld was tissue-specific, because expression was maintained in presomitic mesoderm (Fig. 4 a-d), and dlc expression in the dorsal aorta appeared relatively normal at 22 hpf (Fig. S12, Table S5). Although Wnt16 function is required for somitic expression of dlc and dld, we see no evidence for reciprocal regulation of wnt16 by Dlc and Dld (Fig. S13), or indeed by Notch at all, as expression is unaffected in mib mutants (not shown).

To determine whether alterations in the expression of dlc and/or dld might explain the decrease in HSCs we observe in wnt16 morphant animals, we tested whether loss of function in these genes alone or in combination could alter HSC specification. Beamter (bea) mutants carry a predicted null allele of dlc¹⁸. We compared HSC and T lymphocyte marker expression in wild-type embryos, homozygous bea mutants, embryos injected with dld morpholino (dldMO)¹⁹, and homozygous bea mutants injected with dldMO. We found that runx1 transcripts at 24 hpf, and cmyb at 36 hpf, were greatly reduced in bea homozygous mutants and embryos injected with dldMO (Fig. 3, e-g, i-k: Table S6), but an apparent recovery of HSCs had occurred by 4.5 dpf, when rag1 transcripts, revealing the thymic progeny of HSCs, were readily visible in both groups (Fig. 3 m-o; Table S6). Combinatorial reduction of Dlc and Dld activity, however, eliminated HSCs, as revealed by elimination of runx1 at 24 hpf (Fig. 3, h; Table S6), cmyb at 36 hpf (Fig. 3, l; Table S6), and rag1 transcripts at 4.5 dpf (Fig. 3, p, Table S6). Thus, dlc and dld are combinatorially required for specification of HSCs in zebrafish, and their diminished expression in the somites of wnt16 morphant animals can, in principle, explain the observed loss of HSCs.

To confirm that loss of HSCs in wnt16 morphants is due to loss of dlc and dld, we performed a rescue experiment. We injected embryos with W16MO, and in some cases co-injected mRNA encoding full-length Dlc and Dld ligands singly or in combination. Injection of dlc and dld mRNA did not alone, or in combination have a strong effect on HSC numbers, as measured by runx1⁺ cells in the dorsal aorta (Fig. 3, q-t, y). Co-injection of individual mRNAs with W16MO was unable to restore runx1 in (Fig. 3, u-w, y). However, when injected together, dlc and dld restored runx1⁺ HSCs in a high percentage of W16MO-injected animals (Fig. 3, x, y). These results confirm that decreased dlc and dld expression in wnt16 morphants is responsible for loss of HSCs.

Studies in zebrafish and mouse have shown that Notch specification of HSCs is regulated by Shh and VegfA¹. Shh regulates expression of vegfa, and VegfA signalling is in turn required for vascular expression of Notch receptor genes¹. Notch signalling is required for both arteriovenous and HSC specification^{1, 20-24}. In accord with these observations, loss of Shh signalling causes loss of both artery and HSCs¹. Taken together, one level of control over HSC specification appears to be through a Shh/VegfA/Notch signalling pathway¹. Shh/VegfA/Notch specification of HSCs appears to be distinct from Wnt16/Dlc/Dld effects because vascular and arterial specification is unaffected in W16MO animals (Fig. 2, i-n; Fig. S7, a-d, o, p). Moreover shha and Notch receptor expression are unaffected in W16MO animals (Fig. 2, q, r, Fig. S7, a-l). To confirm these pathways are discrete, we examined expression of six known Shh target genes, with particular interest in vegfaa. None of the Shh targets examined showed significantly altered expression (Fig. S14; Table S7). Our results suggest that the Wnt16 and Shh pathways act in parallel upstream of HSC specification.

The predominant model for how Notch signalling regulates HSC specification is that endothelial cells of the dorsal aorta, receive a requisite Notch1-mediated signal^{1, 21-24}. Chimeric mice generated using Notch1-deleted cells show no contribution of knockout cells to the adult haematopoietic system, demonstrating that Notch1 signalling is required cell-autonomously for specification of HSCs^{1, 21}. The relevant Notch ligand(s) are thought to be expressed in the formed dorsal aorta and/or immediately surrounding mesenchyme^{1, 22-24}. A contributing ligand appears to be Jag1, because knockout mice have severely impaired, but not totally abrogated development of HSCs^{1, 24}. Notch1 activation of Runx1 is thought to occur indirectly through Gata2, since the Runx1 promoter does not contain identifiable Rbpjκ binding sites^{1, 21, 23, 24}. Notch signalling defects in wnt16 morphants suggests that the requirement for Dlc/Dld is not the same as this cell-autonomous requirement, because decreased Notch ligand expression is specific to somites (Fig. 3, a-d), whereas expression of dlc is normal in the dorsal aorta of wnt16 morphants (Fig. S12, Table S5) and dld is not natively expressed near the formed dorsal aorta. Moreover, expression of jag1b and its putative target gata2 are unaffected in Wnt16 knock down animals (Fig. 4, a-d).

To better understand whether Dlc/Dld specification of HSCs work cell-autonomously or not, we sought to define the timing of the Notch signalling event lacking in W16MO-injected animals using animals carrying transgenes that allow heat-shock induction of the Notch intracellular domain (NICD), which is a dominant activator of the Notch pathway^{20, 25}. We induced NICD expression at different time points to identify the temporal window when enforced Notch activity could rescue HSC expression in wnt16 morphants. Nuclear, myc-tagged NICD protein is present in transgenic animals by 1 hour post heat shock (hphs), and is strong from 3 hphs (Fig. 4, e-h) through at least 24 hphs (not shown), consistent with prior reports²⁶. Heat-shock induction of NICD at 14 hpf (10-ss) rescued cmyb expression at 36 hpf (Fig. 4, i-k; Table S8), as has been seen previously for rescue of mib²⁰. Surprisingly, heat shock just two hours later (14-ss) did not rescue cmyb expression along the dosal aorta (red circle, Fig. 4, l, Table S8). These results indicate that the critical phase of Notch signalling required for HSC specification downstream of Wnt16, occurs between 15-17 hpf (12-16-ss) and abruptly terminates by about 18-19 hpf (18-20-ss).

Since this timing is well prior to the formation of the dorsal aorta and HSCs from its ventral endothelium, we sought to determine when cells fated to become HSCs first experience a cell-autonomous Notch signal. We first took advantage of a tp1:Kaede Notch reporter line expressing a green-to-red, photoconvertible Kaede protein under the control of a Notch-responsive promoter. Photoconversion of Notch-responsive, Kaede⁺ cells in the Dlc/Dld-critical window, before 19.5 hpf (21-ss) yielded converted, red fluorescent progeny that contributed to the region near the dorsal aorta at 3 dpf (Fig. 4, m, n), but these cells never became HSCs, as shown by their failure to produce labelled thymic progeny (Fig. 4, m′, n′). In contrast, photoconversion at 24 hpf and later produced both labelled dorsal aorta (Fig. 4, o, p) and thymic immigrants (Fig. 4, o′, p′). Because there is a lag between reception of a Notch signal and the production of mature Kaede protein in the reporter lines, we wanted to confirm when Notch activity appears in pre-haematopoietic cells by a more immediate read-out for Notch-responsiveness. We performed double in situs, in tp1:GFP Notch-reporter transgenic animals, for expression of GFP and etsrp, which labels pre-haematopoietic mesoderm and is required for HSC specification²⁷. The very first double-labelled cells appeared at 18 hpf in the midline (Fig. 4, q, r), and increased in number through at least 24 hpf (Fig. 4, s, t). Taken together, our data suggest that Notch signalling in cells fated to become HSCs begins after 18 hpf and continues well into the second day of development. This timing fits well with the established model of cell-autonomous Notch specification of HSCs, where Notch signalling in nascent HSCs occurs after aortic specification via interactions between Jag1b and Notch1^{1, 21-24}. Overall, our results indicate that somitic signalling by Dlc/Dld downstream of Wnt16 is temporally and spatially distinct from the observable cell-autonomous requirement for Notch signalling. Wnt16 therefore controls a previously unappreciated environmental requirement for Notch signalling in the somites, most simply explained by regulation of an unidentified relay signal.

To better understand what cell population Wnt16-regulated somitic Dlc/Dld acts on, we examined fine somite patterning in Wnt16 and Dlc/Dld loss-of-function animals. Since somite segmentation (Fig. 2, a, b, e, f, Fig. 3, a-d) and myotomal specification (Fig. 2, e, f, Fig. 3, a-d) appear to be intact in wnt16 morphants, we examined specification of the sclerotomal somite compartment, which houses vertebral and vascular smooth muscle cell (VSMC) precursors. Sclerotomal markers displayed severely decreased expression, albeit to variable extents, in both W16MO and combined dlc/dld loss of function animals (Fig. S15; Table S9). Although hypochord (Fig. S15, u-y, green arrowheads), was abolished in the dlc/dld double loss of function animals (Fig. S15, x) as has been reported¹⁸, hypochord was specified normally in wnt16 morphants (Fig. 2, o, p, Fig. S15, y), emphasizing the fact that the Wnt16-dependent loss of somitic dlc/dld is distinct from global loss. Our results suggest that sclerotome specification or morphogenesis is required for HSC specification.

In sum, our results demonstrate that non-canonical signalling by Wnt16 is required genetically upstream of the combined actions of the Notch pathway ligands Dlc and Dld for HSC specification. Dlc/Dld-mediated Notch signalling is spatially and temporally distinct from previously described cell-autonomous requirements for Notch in HSC specification, pointing to the possibility of a novel relay signal. These data represent the first demonstration that non-canonical Wnt signalling activity is required for HSC specification in vertebrates. Given that Wnt16 in mouse is expressed at similar times of development²⁸ and is expressed in embryoid bodies during commitment to blood and vasculature²⁹, it is feasible that this function is conserved in mammals.

1

2

6

7